菏泽学院学报

由于大多数采集到的人脸图像被口罩、帽子、头发遮挡，导致人脸识别效果较差.为此，设计基于循环生成对抗网络的遮挡人脸识别系统.该系统利用电源模块为存储器、人机交互、人脸识别等模块提供稳定电压.系统中的人脸识别模块使用DM6447处理器控制人脸视频图像采集和识别，处理器内置人脸图像自动采集程序，利用该程序控制视频采集前端使用CMOS传感器、可编程逻辑器件等采集人脸视频图像，并基于循环生成对抗网络的遮挡人脸识别算法识别遮挡人脸图像后，通过通信接口模块将识别结果传输到存储器模块和人机交互模块分别进行存储和展示.实验结果表明：该系统具备良好的运行稳定性、通信性能以及图像缩放功能，且可有效识别不同类型遮挡的人脸，识别精度高达98%.

Most of the collected face images are blocked by masks, hats or hair, resulting in poor face recognition. For this reason, this paper designs an occluded face recognition system based on cyclic generation antagonism network. The system uses power supply device to provide stable voltage for memory, human-computer interaction, face recognition and other modules. The face recognition module uses Processor DM6447 to control the image acquisition and recognition of the video image. The processor has a built-in face image automatic acquisition program, uses CMOS sensors, programmable logic devices, etc. to collect the face video and image, uses the occluded face recognition algorithm based on the circularly generative confrontation network to identify the occluded face image, and then transmits the recognition results to the memory and human-computer interaction module through the communication interface module for storing and displaying separately. The experimental results show that the system has good operation stability, communication performance and image scaling function, and can effectively recognize different types of occluded faces, with recognition accuracy of 98%.